Linear vs nonlinear for estimation!

Dynare help (legacy posts)
1

Just generally speaking. Let’s take a simple new keynesian model with sticky prices and sticky wages, such as the one in Gali (2008) in chapter 3. Will there be a difference between log linearizing by hand or letting dynare linearize everything when estimating?

Please note, when letting Dynare linearizing, price and wage setting must be rewritten in differential equations since it contains infinite sums.

So what’s the general verdict for estimation? Will it give identical results or is there an advantage with either method?

2

Apart from potentially different observation equations due to manual loglinearization typically also involving demeaning, the two are identical and yield identical estimation results. Thus, there is no need in doing manual loglinearization.

3

Oki, thanks!